Do we need repeated simulations to study the effect of random fabrication tolerances on a photonic circuit?
The obvious answer would be yes. Of course we need many simulations if we use Monte Carlo. Even if we use more complex techniques to represent uncertainty we still normally need to run several simulations to sample the device random response. But what if we try to subvert this approach?
In other words, can we can obtain a full description of the behaviour of a circuit under fabrication tolerances with a single deterministic simulation, saving time and computational resources?
This is actually what we demonstrate in our latest work where we exploit a generalised polynomial chaos expansion approach to realise a completely novel class of device models to be used within photonic Process Design Kits. These stochastic models inherently convey stochastic information, are circuit independent and can replace the original deterministic models in the design kit. They can hence be combined according to the building blocks connections to derive with a single run of a deterministic simulator the stochastic behaviour of any circuit, enabling an unprecedented simulation efficiency.